JustCall

The tier structure settles the removal test without any argument being needed. Model driven capability is packaged and sold separately rather than built in: the review and coaching layer is a nine dollar per user add on at the two lower tiers, and the voice agent is an entirely separate product with its own three pricing plans and its own per minute rate.

Strip both away and the entry tier at twenty nine dollars is still a complete cloud telephone system, with call menus, routing, distribution, recording, voicemail, messaging, numbers in more than seventy countries, over a hundred record system integrations, analytics and programmatic access. That is what most of this vendor's customers are buying and it contains no inference at all.

The model layer improves a working product rather than constituting it, which is the established platform pattern this index grades in the middle of the band. Ask what the platform costs and delivers with the review add on and the voice agent both excluded.

A voice agent that answers and places real telephone calls, with a genuine escalation path and little else published. The autonomy is substantial and stated: the agent runs around the clock, and its published action set lets it collect data, schedule appointments, send messages, assign disposition codes and transfer calls, with custom actions and outbound calling at the upper plan, the latter marked as in testing.

Transferring to a person is named as a first class action, which is the escalation route this axis looks for and is more than many agent products publish. Around it sit advanced roles and permissions from the second tier and knowledge bases scoped per account.

What is not published is anything bounding behaviour: no statement of what the agent may commit to, when it must hand over, whether conversations are reviewed, what happens when it books an appointment wrongly, or what limits apply to outbound campaigns run by an agent rather than a person. Ask what the agent may not do, and what triggers a mandatory transfer to a person.

The interfaces and the metering are specific, the model layer is not disclosed. On the published side a developer reference is named and linked for programmatic access and webhooks, transcription overage is priced at a stated rate per minute, the voice agent's language support is given as a count, its included minutes are stated per plan, and its capability set is enumerated action by action. A buyer can establish what is bought and what it costs to exceed.

On the undisclosed side nothing at all is named for transcription, scoring, sentiment, script checking, agent assist or the voice agent itself, and no provider, model family or version appears anywhere reached. Voice cloning is offered at the top agent plan with no description of how a voice is captured, verified as belonging to the person cloned, or retired. Ask which models power transcription, scoring and the voice agent, and what consent and verification govern a cloned voice.

Three customer results, each attached to a named person at a named company with a case study behind it, which is the form this axis rewards. A lighting manufacturer reports dropped calls falling by thirty to forty percent, quoted by its chief growth officer by name. An education company reports a twenty five percent rise in agent productivity and around ten hours saved per account manager per week, attributed to its global operations manager.

A pharmacy reports engagement up more than twenty percent, attributed to its chief executive. Each links to a full case study, and a broader customer story section, a customer review wall and named logos across consulting, media and education sit behind them, alongside independent review standing across a large review base. It is held below the top band by the measurement basis, absent in all three cases: no period, no baseline, no method and no definition of the metrics. A separate claim of powering more than a billion conversations carries no date or counting basis at all. Ask over what period each figure was measured and against what baseline.

Registration obligations are handled and the dialing regulation this product most engages is never mentioned. On the credit side, messaging registration for the United States is acknowledged as a real cost with compliance charges described in the help material, script checking is shipped as a named feature at the upper tiers, a number health capability exists on the outbound package, and the fair usage policy sets published numeric ceilings.

Against that, the product sells a predictive dialer running up to ten lines per agent, which is the architecture that abandoned call regulation exists to govern, and nothing on any page reached states an abandonment rate ceiling, an answering machine detection policy, a calling window restriction or a screening step against do not call registers.

Two records in this lane hold the top grade for exactly those things, one for engaging the regulation by name and one for putting register screening in its cheapest plan. Ask what abandonment rate the predictive dialer enforces and whether numbers are screened against do not call registers before dialing.

The instruments are published, were not read, and the data protection terms are sold rather than provided. A privacy policy, terms and conditions, a responsible disclosure policy and a trust centre are all published, and none was retrievable on this pass, so retention periods, lawful bases and the rights exercise route are unestablished and nothing here asserts they are absent. What the pricing page does establish is a gating decision worth recording.

Enterprise grade compliance covering the health information regime and European data protection is listed as a feature of the quoted tier with a ten licence minimum, which places contractual data protection terms behind a negotiated purchase rather than offering them to every customer. That matters because this product records telephone calls and stores message content involving people who are not customers, on every tier. Ask what data protection terms apply on the published tiers, and what the default retention period is for call recordings and message history.

No purchased corpus, and the material at stake arrives from the customer and the network. There is no contact database, no enrichment and nothing brokered here: the numbers dialed come from the customer's own record system, and what the platform accumulates is call recordings, transcripts and message history generated by use.

So the usual chain of sources, licences and resale permissions does not apply and most of the axis is out of scope, which is why this sits in the middle of the band with the scope stated. What remains in scope is that recordings and transcripts contain the speech of the people called, who are not customers of either party and whose consent to recording is a legal precondition in many jurisdictions.

Nothing published describes how the platform supports, prompts for or evidences that consent, or whether an announcement can be enforced at the account level. Ask whether a recording announcement can be made mandatory per jurisdiction and what consent record is retained.

The telephony is the vendor's own and the software connections run through sanctioned routes. Numbers are provisioned by the vendor in more than seventy countries and calls and messages traverse carrier infrastructure the vendor contracts for, so a customer is not borrowing access to a platform that could withdraw it, and porting numbers in and out is offered as a supported process.

The integration surface reaches customer record, support and sales engagement systems through each product's documented route, with more than a hundred named and a public developer reference for anything custom, plus mobile applications distributed through both official stores. That architecture is what this axis rewards.

It stops below the top band because the governing regime here is carrier and messaging registry rules rather than software platform terms, and while messaging registration is acknowledged as a cost, no conformance position of the vendor's own is stated against carrier requirements or the consequences of a campaign registration being revoked.

The load bearing commitment for a product of this shape is not stated anywhere reached. This platform ingests recorded telephone calls and message content and runs transcription, scoring, sentiment analysis and coaching over them, which makes the training question direct: do a customer's recorded conversations train or tune the models, and does a third party model provider retain them.

No statement on either appears on the pricing page, in the questions section or in the navigation, where a comparable vendor graded this session places its position on its own pricing page. The trust centre was not retrievable on this pass, so nothing here asserts the commitment is absent, only that it was not found where a buyer would look. Voice cloning at the top agent plan raises a second unaddressed question about what is retained of a cloned voice and who may use it. Ask for a written statement that recordings do not train any model, and what a cloned voice sample is retained for and for how long.

A synthetic voice can place calls under a cloned likeness and nothing published tells the person answering. The upper voice agent plan lists voice cloning and outbound calling together, so the product supports placing telephone calls in a voice modelled on a real person to recipients who did not initiate contact, and no page reached states that the agent identifies itself as not human, that an announcement is made, that disclosure can be enforced at account level, or that any jurisdiction specific handling exists.

That is the central failure this axis records, and it sits on a telephony product where the use of artificial voices in calls is an active regulatory question rather than an etiquette one. The same silence covers call recording: no consent prompt, announcement setting or consent record is described anywhere for a platform that records conversations by default across every tier.

Graded at the floor because no disclosure mechanism of any kind was located, not because one was found and judged thin. Ask what a person answering an agent call is told, and what recording announcement the platform can enforce.

Programmatic access on the entry tier and a stated count backed by named destinations. More than a hundred integrations are claimed and fifteen are named individually with their own pages, spanning three customer record platforms, two support desks, three sales engagement tools, a scheduling product, a messaging platform, a prospecting database and a general automation connector, so the claim is checkable rather than a logo wall.

Interface access and workflows appear in the entry plan's feature list rather than being reserved for the largest customers, with higher limits at the second tier and maximum limits at the quoted tier, and a public developer reference is named and linked in the questions section. Mobile applications ship on both official stores, and an integration with a vehicle interface exists.

It stops below the top band because the developer reference was not retrieved on this pass so what the interface exposes is unverified, and because rate limits are tiered in a way the page does not quantify at any level.

Telephony geography is documented in unusual detail and data geography is not addressed. The calling side is specific: numbers in more than seventy countries, a region selector on the pricing page covering around eighty countries that changes the published rates, local number bundles available in seven named markets, and per region inclusion and exclusion lists stating exactly which call routes and message destinations fall inside a bundle.

A buyer can establish where they can call and what it costs with more precision than almost any record in this lane. What is absent is where anything is stored. No hosting region, processing location, residency election or cross border transfer mechanism for recordings, transcripts or message history appears on any page reached, and the trust centre was not retrievable.

That gap matters because the stored material is recorded speech from whichever of those eighty countries a customer dials. Ask in which regions recordings and transcripts are stored and whether a region can be elected.

A substantial certification set displayed on the page, a live trust centre behind it, and the vendor contradicting itself about what it actually holds. Six compliance marks appear in the security section covering a service organization control attestation, the health information regime, the international information security standard, the payment card standard, European data protection and a cloud security alliance registry level, and a trust centre is linked from the quoted tier's service level commitment.

That is a broader set than most records in this lane publish. The contradiction is specific and matters: the questions section on this same page states that the parent company has received an attestation of the point in time type, while other pages on this vendor's own site describe the period type, and those are materially different assurances.

Held below the top band on that inconsistency, on the trust centre not being retrievable on this pass, and on no auditor, audit period or report request route being named. Ask which attestation type is current, its audit period and the auditor.

Unusually complete usage disclosure sitting alongside a plan card that contradicts it. The strong parts are genuinely strong: three tiers carry per user figures, licence minimums are stated on every plan, prices change with a selectable calling region across around eighty countries, the coaching add on is priced per user, the voice agent carries its own three plans and a per minute rate, local number bundles are priced at three volumes, transcription overage is given as a rate per minute, and a fair usage policy is published in full with per region inclusion and exclusion lists.

Against that, the plan cards advertise unlimited outbound and inbound minutes while the fair usage tables on the same page give a thousand minutes per user for each, which is the number that will actually govern the bill. Two tiers are quoted, the messaging registration fees are documented only in the help material rather than on the pricing page, and one number bundle popup shows a struck through figure lower than the price beside it. Ask which minute allowance the contract carries and what messaging registration costs on top.

The asset most likely to lock a customer in can be taken away, and the records built on it are undescribed. Telephone numbers port both in and out, with the vendor stating it handles inbound porting without service interruption and publishing porting fees as a real cost, which is the single most important exit right for a telephone system because a business number is an operational dependency rather than a data file. Programmatic access on every tier provides a technical retrieval path.

What is absent is everything about the data. No export route is described for call recordings, transcripts, message history, scoring or analytics, and no statement exists on what happens to any of it after cancellation or how long a departing customer has to retrieve it. The one published deletion statement concerns lapsed trials, where the account is stated to be deleted thirty days after the trial ends. Ask for a documented export of recordings, transcripts and message history, and the retention window after a paying account closes.

Real mechanisms for the voice equivalent of deliverability, and no measurement of any of them. What is published is more than most in this lane: a number health capability on the outbound package, local number bundles in seven markets sold on the reasoning that a local presence improves answer rates, and a stated allowance of one free swap per month for numbers that are underperforming, which is a direct and practical answer to numbers being flagged or spam labelled by carriers.

Messaging registration for the United States is acknowledged as a precondition with fees. Fair usage ceilings are published as numbers per user per region. Against that, the vendor claims industry leading connect rates with no figure, no answer rate, no spam labelling incidence and no message delivery rate is published anywhere, and nothing describes how the predictive dialer paces against answer detection. Ask for measured connect and answer rates by region, and how often numbers are flagged and swapped.

Coverage documented across three dimensions rather than asserted on one. Seven industries carry their own pages, spanning healthcare, education, solar, fitness and wellness, real estate, financial services and retail, and three functional buyers are addressed separately across sales, support and technology operations, which is broader vertical treatment than any record graded in this batch.

Geographic reach is evidenced concretely rather than claimed, with numbers in more than seventy countries, a pricing region selector covering around eighty, and local number bundles named for seven specific markets. The commercial ladder runs from a two licence entry tier through two further published tiers to two quoted arrangements, one of them requiring ten licences, so small teams and large operations both have a described path.

It stops below the top band because no company size guidance accompanies the tiers beyond licence minimums, and because support coverage and service hours are not stated for any region outside the quoted tier's dedicated line.